Learning strategy for the control of a one-legged hopping robot

A study is made of neural network strategies for the control of a dynamic, locomotive system, using as a model a one-legged hopping robot. The control task is to make corrections to the motion of the robot that serve to maintain a fixed level of energy (and minimize energy losses), which yields a stable periodic limit cycle in the system's state space corresponding to periodic hopping to a prespecified height. The studied models are D. Michie and R. A. Chamber's BOXES system (1962), the ASE/ACE configuration of A. G. Barto et al. (1983), and the Anderson-Sutton two-layered connectionist model. Results are demonstrated through numerical simulations and are quantitatively compared to the performance obtained by M. H. Raibert (1984) for the robotic leg, using full-state feedback. The main difference between Raibert's solution and the neural strategies presented here is that the authors' system is not aware of the dynamical model of the plant that it is to control.